Google’s New Google Labs AI Tools Update (STAX, Flow, Mixboard, and More)

Google Labs has quietly become one of the easiest places to try Google’s newest AI ideas without installing anything. It’s basically a public sandbox where Google ships experiments, watches how people use them, and improves (or retires) them fast.

That speed matters right now. In December 2025, a lot of creators and small teams are stuck in the same loop: write the copy, design the assets, rebuild the slides, then scramble to turn it into video. Google’s latest Labs updates point at a different workflow, where one browser tab can help you move from rough idea to polished output in a fraction of the time, and where you can actually test which model gives you the best results instead of guessing which one “feels smartest.” If you’ve been collecting ai tools but not getting consistent output, this update is worth your attention.

What is Google Labs, and how do you access these new AI tools?

Google Labs (find it at https://labs.google) is Google’s “try it early” hub. It’s a collection of experimental features that often show up here before they become polished products inside Workspace, Gemini, or other Google apps.

A few expectations to keep you sane:

• Experiments change quickly. Buttons move, names change, and features can appear or disappear.
• Access can vary. Some tools are limited by region, account type, age settings, or waitlists.
• Quality can swing. These are prototypes. You’ll get moments of “wow” and moments of “why did it do that?”

Still, the barrier to entry is low. Most of the time you sign in, pick an experiment, and you’re creating in minutes.

Quick setup checklist to try Google Labs in under 5 minutes

• Sign in to your Google account at https://labs.google
• Tap View all to see the full list of experiments
• Pick one tool that matches today’s task (slides, video, planning, or evaluation)
• Run the default example once, just to see the expected input format
• Try one prompt tied to your real work (a product, a client, a channel, a deck)
• Save the best output and the prompt that produced it

One safety note: don’t paste private info (client data, personal IDs, passwords, contracts) into experimental tools.

The standout tools in the new Google Labs update (and what they are best at)

This update isn’t just “more AI.” It’s more specific than that. Google is filling obvious gaps in modern workflows:

• Evaluation: stop arguing about which model is best and measure it.
• Creative production: generate visuals, slides, and video in a tight loop.
• Brand consistency: reduce the “every post looks different” problem.
• Planning: give messy ideas a home before you turn them into deliverables.

A good overview of one of the biggest upgrades (Mixboard plus Nano Banana Pro) is covered in Google’s own post, Create presentations with Nano Banana Pro in Mixboard and more.

STAX (stacks.withgoogle.com): compare models, score outputs, and pick the best one

STAX (https://stacks.withgoogle.com) is the tool that makes the rest of your workflow smarter. In plain English, it lets you run the same prompt against different models and compare outputs side by side.

If you’ve ever thought, “Gemini feels better for this,” or “OpenAI writes cleaner hooks,” STAX turns that vibe into something you can actually validate.

What you can do inside STAX:

• Prompt playground: paste the same system instructions and user prompt, then run multiple models.
• Model comparison: switch between providers and versions (Gemini, OpenAI, Claude, and others supported in the selector list).
• Evaluation and scoring: rate outputs on things like instruction-following, fluency, groundedness, chat quality, and safety.
• Practical metrics: track latency (how fast it responds), tokens, and run history so you can see what changed.

A simple use case that mirrors real creator work: You want YouTube packaging that fits your style, not generic “Top 10” titles.

Try something like:

• “Generate 10 YouTube titles for a video about Google Labs updates, keep them under 55 characters, no clickbait words.”
• “Now write 6 opening hooks in a casual tone, each under 12 seconds spoken.”

Run it on two models, then score the results. After a few rounds, you stop guessing and start building a shortlist of “best model for titles,” “best model for hooks,” and “best model for safe rewrites.”

This isn’t only for people building AI products. It’s useful for everyday work because it reduces the biggest hidden time-waster in AI: rerolling prompts because you picked the wrong model first.

If you want more context on why model choice is becoming a real strategy (not a nerd detail), this internal breakdown is a solid companion read: GPT‑5.1 vs Gemini 3 Pro: AI battle overview.

Flow, Mixboard, Pamelli, and “Beautify slide”: create videos, visuals, and on brand content faster

These tools feel like they were designed for the reality of modern work: you rarely need just one thing. You need a concept, visuals, a deck, and something short-form for social.

Flow (labs.google/fx/tools/flow) focuses on video generation. The most practical feature is the ability to create a short clip, then add to scene to extend it. That matters because many AI video tools produce a nice 6 to 8-second moment, then fall apart when you ask for a full story. Flow nudges you toward building longer content in small, controlled chunks.

Example task:

• “Create an 8-second product teaser video of a new app interface on a phone, bright studio lighting, smooth camera movement.”
• Then: “Add a second scene showing a person using it in a coffee shop.”

Mixboard is best described as a Pinterest-like idea board with an infinite canvas. You can drop in images, remix styles, plan a project, and keep everything visible at once. It’s great for moodboarding campaigns, thumbnails, room redesign ideas, even event planning. Publications noticed this shift toward turning “idea dumps” into finished presentations, including Digital Trends’ coverage of Mixboard’s presentation feature.

Example task:

• “Build a moodboard for an 80s Miami New Year’s living room party, neon accents, chrome decor, tropical textures.”

Pamelli is about on-brand campaigns. The standout idea is “business DNA,” you provide your site (and sometimes assets), and it pulls signals like colors, fonts, taglines, and brand values. Then it generates campaign visuals and copy that stay consistent, instead of looking like five different freelancers guessed your brand vibe.

Example task:

• “Create an Instagram campaign for my product that helps people go viral on TikTok, include 3 ad concepts, headline options, and a short CTA.”

Google Slides: “Beautify slide” is the most instantly useful feature for normal people. If you’ve ever made a slide that looks like a ransom note of mismatched fonts, this button is relief. It takes an ugly slide and proposes cleaner layouts, visuals, and hierarchy, and you can regenerate variations until one fits.

Example task:

• Take a busy slide with 7 bullets and a low-quality image, then use Beautify to produce 2 to 3 cleaner options, pick the best, and edit the text.

For more detail on Mixboard’s Nano Banana Pro upgrade from another angle, Chrome Unboxed’s write-up is a useful quick scan.

 Google Labs-style homepage on a laptop, showing a grid of experimental tools, created with AI.

How to use these tools together for real work (content, marketing, and simple automation)

The real win is combining them so each tool does what it’s best at. Think of it like a small team:

• Mixboard is your planner and visual memory.
• Pamelli is your brand-aware campaign generator.
• Beautify slide is your fast designer for decks.
• Flow is your video editor for short promos.
• STAX is your quality control and model picker.

This is where collecting ai tools starts to feel less like a hobby and more like a system you can repeat weekly.

For creators, this combo helps you ship packages for YouTube, Instagram, and TikTok without spending half your life on formatting.

For small businesses, it helps you turn one offer into multiple assets (ad creative, short promo video, and a deck for partners) without hiring a full production crew for every launch.

If you’re curious how Google is positioning Nano Banana Pro for creators across these experiments, this internal guide adds extra background: Explore Nano Banana Pro’s 8 hidden features.

A beginner friendly workflow you can copy: plan, generate, polish, publish

1. Plan in Mixboard
   Prompt: “Create a moodboard for a campaign that helps a creator go viral on Instagram and TikTok, bold colors, clean typography, energetic vibe.”
2. Generate campaign assets in Pamelli
   Prompt: “Using my brand style, generate 3 ad creatives and 5 headline options focused on: ‘go viral faster with less effort.’ Keep the tone confident, not hypey.”
3. Polish the pitch in Google Slides (Beautify slide)
   Prompt: “Rewrite this slide headline to be clearer and shorter: ‘We help creators scale content output with AI automation.’ Give 5 options under 8 words.”
   Then run Beautify on the slide to clean layout and hierarchy.
4. Publish short video variants in Flow
   Prompt: “Turn this campaign hero image into an 8-second vertical promo clip, smooth motion, subtle light shifts.”
   Follow-up: “Add a second scene with a phone screen showing the before and after results.”
5. Improve the prompts and pick models in STAX
   Run your best prompts on two models. Keep the one that hits your tone with fewer edits. Track latency if speed matters for batch work.

Want a third-party summary of Mixboard’s presentation generation feature in a mobile context? Android Central’s overview is helpful.

What I learned after testing Google’s new tools (wins, limits, and smart tips)

The biggest win surprised me: comparison beats confidence. I used to assume one model was “my model” for most tasks. After running the same prompt through STAX and scoring outputs, the pattern was obvious. Some models are better at tight structure, some are better at tone, and some are simply faster. Picking the right one up front saved me more time than rewriting prompts.

I also learned that “on brand” isn’t a nice-to-have. When Pamelli has enough input (site, voice, offer, and basic brand rules), the output stops feeling random. Without that, you get the classic AI problem: decent content that still doesn’t feel like you.

Flow taught me a practical lesson about AI video: don’t ask for a full story in one go. Build in small scenes, then extend. When I treated it like directing short shots, the results stayed coherent longer.

A few smart habits made everything better:

• Keep prompts short unless you need strict constraints.
• When you need precision, write an expected output (format, length, and what “good” looks like).
• Save your best prompts, not just your best outputs.
• Watch for confident errors. Even polished outputs can include wrong facts, odd claims, or mismatched details.

There’s also a quiet pressure behind all of this. AI keeps raising expectations for speed and output at work. Some forecasts say a huge number of roles will be affected over time, but the practical takeaway is simple: learn the tools, build repeatable workflows, and make yourself the person who can ship quality fast.

Conclusion

Google Labs is turning into a serious testing ground for creation, evaluation, and brand-aware content generation. The smart approach isn’t trying everything at once, it’s doing one small real project and keeping what works.

Pick one tool today (STAX, Flow, Mixboard, Pamelli, or Beautify slide), run a short task you already need to do, and keep a folder of your best prompts and outputs. That habit turns curiosity into momentum, and momentum is what actually changes your work.